CN105740882A - Target identification method and target identification device based on multi-scale invariant description - Google Patents

Abstract

The invention discloses a target identification method and a target identification device based on multi-scale invariant description. The method comprises the following steps: first, getting the shape of a to-be-identified target as a target shape, extracting a closed contour from the edge of the target shape, and getting the contour points and the coordinate of each contour point to complete extraction of the contour information of the target shape; then, calculating the parameters (namely, area parameter, arc length parameter and center of gravity parameter) of each contour point in each contour layer as a complete multi-scale invariant descriptor of the contour point to complete extraction and effective expression of the global features and local features; and finally, matching the to-be-identified target with templates based on the multi-scale invariant descriptor of each contour point in the target shape to get the optimal matching template corresponding to the to-be-identified target so as to determine the category of the to-be-identified target. Thus, extraction and effective expression of the global features and local features of the target shape are realized, and high identification efficiency and high identification accuracy are both ensured.

Description

A method and device for target recognition based on multi-scale invariant description

technical field

The present invention relates to the technical field of target recognition, and more specifically, to a method and device for target recognition based on multi-scale invariant description.

Background technique

Machine vision cognition has always been a research hotspot. Using object shape features for target recognition is the main research topic of machine vision. The latest progress in this research is to design intelligent shape descriptors to fully extract target shape features for more accurate A good similarity measure has been widely used in engineering, such as wide baseline matching, target recognition and classification, image and video matching and retrieval, robot automatic navigation, scene classification, texture recognition and data mining and other fields.

Generally, object recognition methods based on shape description are divided into two categories according to the source of features: contour-based methods and transform domain-based methods. Among them, the former features all come from the target contour, such as Moravec, Harris corner feature, contour perimeter, compactness, eccentricity, Hausdroff distance, etc., which are simple but effective, so they have been widely used in the field of robot vision. . Contour-based methods mainly describe target features in two ways: based on global features and based on local features. Among them, based on global features, it can describe the overall characteristics of the target, which is especially useful for contours with simple target shapes and few local details, but it is not sensitive to local shape changes, and the detail discrimination is not high, such as ShapeContexts, Inner-Distance and Multi-scaleRepresentation etc. Therefore, the recognition accuracy is low. However, based on local features, the above-mentioned problem of local discrimination can be overcome, and even if part of the target contour is occluded or deformed, other local features can also be matched and recognized, but due to the loss of global shape structure information, the recognition accuracy is still affected, such as ShapeTree , ClassSegmentSets, ContourFlexibility, etc. Moreover, the calculation complexity of the above methods is high, so the recognition efficiency is low.

To sum up, how to provide an object recognition method capable of ensuring high recognition efficiency and recognition accuracy at the same time is an urgent problem to be solved by those skilled in the art.

Contents of the invention

The purpose of the present invention is to provide a target recognition method and device based on multi-scale invariant description, so as to ensure high recognition efficiency and recognition accuracy at the same time.

In order to achieve the above object, the present invention provides the following technical solutions:

A target recognition method based on multi-scale invariant description, including:

Obtaining the shape of the target to be identified as the target shape, and extracting a closed contour from the edge of the target shape, obtaining all contour points on the contour and the coordinates of each contour point;

Determining the number of layers of the contour, and calculating the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to each layer based on the coordinates of each contour point, as the multi-scale invariant descriptor of the contour point;

Using the multi-scale invariant descriptor of each contour point, match the target to be recognized with the template in the preset template library to obtain the best matching template of the target to be recognized, and determine the best The category of the matching template is the category of the target to be identified.

Preferably, a closed contour is extracted from the edge of the target shape, including:

A Canny operator is used to extract a closed contour for the edge of the target shape.

Preferably, determining the number of layers of the outline includes:

Step A: Determine that the current layer is 1;

Step B: add 1 to the current layer as the current layer, calculate the area parameter, arc length parameter and center of gravity parameter corresponding to each of the contour points corresponding to the current layer and the area parameter of the layer obtained by subtracting 1 from the contour point corresponding to the current layer , the difference between the arc length parameter and the center of gravity parameter, and judge whether the ratio of the sum of the differences corresponding to all the contour points to the number of the contour points is less than the difference threshold, if so, then determine the current layer minus 1 to obtain The layer number corresponding to the layer is the layer number of the outline, if not, then perform step C;

Step C: Go back to step B.

Preferably, calculating the area parameter, arc length parameter and center of gravity parameter corresponding to each of the contour points of the current layer includes:

Determining any contour point as a target contour point, taking the coordinates of the target contour point as the center, and making a circle with a radius corresponding to the current layer as a preset radius, to obtain a preset circle corresponding to the current layer;

Taking the ratio of the area of the area intercepted by the preset circle in the target shape and having a direct connection relationship with the target contour point to the area of the preset circle as the area parameter of the target contour point;

Taking the ratio of the length of the arc segment cut out by the preset circle in the target shape and having a direct connection relationship with the target contour point to the circumference of the preset circle as the arc of the target contour point long parameter;

Determining the distance between the center of gravity of the region intercepted by the preset circle in the target shape and having a direct connection relationship with the target contour point and the target contour point, and calculating the ratio of the distance to the preset radius As the center of gravity parameter of the target contour point.

Preferably, determining the radius corresponding to the current layer as a preset radius includes:

The ratio of the equivalent radius of the target shape to 2 to the Nth power is used as the preset radius corresponding to the current layer, where N is the number of layers corresponding to the current layer.

Preferably, determining the equivalent radius of the target shape includes:

calculating the area of the target shape, and taking the square root of the area of the target shape to obtain an equivalent radius of the target shape.

Preferably, using the multi-scale invariant descriptor of each of the contour points, the target to be identified is matched with templates in a preset template library to obtain the best matching template of the target to be identified, including:

Calculating the matching degree between the multi-scale invariant descriptor of the target to be identified and the multi-scale invariant descriptor of the template, and determining the template whose matching degree is not greater than that of other templates as the best matching template .

Preferably, the calculation of the matching degree between the multi-scale invariant descriptor of the target to be identified and the multi-scale invariant descriptor of the template includes:

Arranging the contour points of the target to be identified in order to form a target sequence, and arranging the contour points of a template that needs to be matched with the target to be identified in order to form a matching sequence;

Using a dynamic programming algorithm to calculate the matching degree between the target sequence and the matching sequence as the matching degree between the target to be identified and the corresponding template.

Preferably, using a dynamic programming algorithm to calculate the matching degree between the target sequence and the matching sequence includes:

Calculate the Euclidean distance between the multi-scale invariant descriptors corresponding to every two contour points, and determine the distance as the matching cost between the two contour points, wherein the two contour points are respectively included in said target sequence and said matching sequence;

The sum of the obtained matching costs is used as the matching degree.

A target recognition device based on multi-scale invariant description, comprising:

The extraction module is used to obtain the shape of the target to be identified as the target shape, and extract a closed contour from the edge of the target shape, and obtain all contour points on the contour and the coordinates of each contour point;

Determining the number of layers of the contour, and calculating the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to each layer based on the coordinates of each contour point, as the multi-scale invariant descriptor of the contour point;

Using the multi-scale invariant descriptor of each contour point, match the target to be recognized with the template in the preset template library to obtain the best matching template of the target to be recognized, and determine the best The category of the matching template is the category of the target to be identified.

A target recognition method and device based on multi-scale invariant description provided by the present invention first obtains the target shape of the target to be recognized, then extracts the closed contour from the edge of the target shape and obtains all contour points and the coordinates of each contour point , to achieve the extraction of the contour information of the target shape; then calculate the parameters of each contour point in each contour layer, that is, the area parameter, arc length parameter and center of gravity parameter, as a complete multi-scale invariant descriptor for each contour point , in order to realize the extraction and effective representation of global features and local features, and finally match the target to be recognized with the template according to the multi-scale invariant descriptor of each contour point in the target shape, and obtain the best matching template corresponding to the target to be recognized , to determine the category of the target to be identified. Therefore, in the above process, not only focus on global features or local features, but also consider global features, local features, and the relationship between global features and local features at the same time, and conduct multi-scale, multi-level, and multi-faceted analysis , or in other words, through the above steps, the extraction and effective representation of the global and local features of the target shape are realized, thus avoiding the low recognition accuracy in the background technology based only on global or local features; and, different Due to the high computational complexity of the target recognition method provided by the background technology, the multi-scale invariant descriptor used in the embodiment of the present invention when matching the target to be recognized with the template usually has a small dimension, so the computational complexity is relatively small. Low, achieving high recognition efficiency and recognition accuracy at the same time.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only It is an embodiment of the present invention, and those skilled in the art can also obtain other drawings according to the provided drawings without creative work.

FIG. 1 is a flow chart of an object recognition method based on multi-scale invariant description provided by an embodiment of the present invention;

FIG. 2 is a schematic diagram of a specific example of the target shape involved in a target recognition method based on multi-scale invariant description provided by an embodiment of the present invention;

Fig. 3 is a flow chart of determining the number of layers of a contour in an object recognition method based on multi-scale invariant description provided by an embodiment of the present invention;

FIG. 4 is a specific schematic diagram of a target shape in a target recognition method based on multi-scale invariant description provided by an embodiment of the present invention;

5 is a schematic diagram of a target shape intercepted by a preset circle in a target recognition method based on multi-scale invariant description provided by an embodiment of the present invention;

Fig. 6 is a schematic diagram of a target recognition method based on multi-scale invariant description provided by an embodiment of the present invention after the target shape is divided by a preset circle;

FIG. 7 is a schematic structural diagram of an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention.

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

Please refer to Figure 1, which is a flow chart of a target recognition method based on multi-scale invariant description provided by an embodiment of the present invention. It should be noted that the target recognition method based on multi-scale invariant description provided by an embodiment of the present invention The technical fields to which the method can be applied include but are not limited to building recognition, vehicle recognition, gesture recognition and commodity classification, etc., and may specifically include the following steps:

S1: Obtain the shape of the target to be recognized as the target shape, and extract a closed contour from the edge of the target shape, and obtain all contour points on the contour and the coordinates of each contour point.

It should be noted that the target shapes involved in the present invention can all be shapes with closed contours, as shown in FIG. 2 , which is a specific example of the target shapes involved in the present invention. In addition, the number of contour points is the number of all points on the contour, and its specific value is determined according to the actual situation, and the contour features that completely represent the target shape shall prevail.

In a digital image, the edge of a shape can be represented by a series of contour points with coordinate information, and the set S of contour points of the target shape in the embodiment of the present invention can be expressed as:

S={p(i)|i∈[1,n]}

Among them, n represents the length of the contour, that is, the number of contour points; p(i) represents the i-th contour point in the sequence of contour points, and has:

p(i)={u(i),v(i)}

Among them, u(i), v(i) are the horizontal and vertical coordinates of p(i), respectively.

S2: Determine the number of layers of the contour, and calculate the area parameter, arc length parameter, and center of gravity parameter of each contour point corresponding to each layer based on the coordinates of each contour point, as the multi-scale invariant descriptor of the contour point.

It should be noted that the number of layers of the contour is the number of layers of the multi-scale invariant descriptor corresponding to each contour point. By determining the number of layers of the contour, the area parameter of each contour point under the contour of each layer can be calculated, Arc length parameters and center of gravity parameters to form a complete multi-scale invariant descriptor for each contour point. Wherein, the number of layers of contour points can be determined by the staff according to actual needs. If the number of layers can be set directly, it can also be obtained according to a certain algorithm, all of which are within the protection scope of the present invention.

S3: Using the multi-scale invariant descriptor of each contour point, match the target to be recognized with the template in the preset template library, obtain the best matching template of the target to be recognized, and determine the category of the best matching template as the target to be recognized Identify the class of objects.

Among them, the preset template library is a template library preset by the staff, each template in the preset template library has a corresponding multi-scale invariant descriptor, and the calculation method of the multi-scale invariant descriptor of each template The calculation method is the same as that of the multi-scale invariant descriptor of the target to be recognized, and will not be repeated here. Match the multi-scale invariant descriptor of each contour point of the target shape with the multi-scale invariant descriptor of the contour point of each template, and then obtain the best matching template matching the target to be recognized, and determine the best The category of the matching template is the category of the target to be recognized.

A target recognition method and device based on multi-scale invariant description provided by the present invention first obtains the target shape of the target to be recognized, then extracts the closed contour from the edge of the target shape and obtains all contour points and the coordinates of each contour point , to achieve the extraction of the contour information of the target shape; then calculate the parameters of each contour point in each contour layer, that is, the area parameter, arc length parameter and center of gravity parameter, as a complete multi-scale invariant descriptor for each contour point , in order to realize the extraction and effective representation of global features and local features, and finally match the target to be recognized with the template according to the multi-scale invariant descriptor of each contour point in the target shape, and obtain the best matching template corresponding to the target to be recognized , to determine the category of the target to be identified. Therefore, in the above process, not only focus on global features or local features, but also consider global features, local features, and the relationship between global features and local features at the same time, and conduct multi-scale, multi-level, and multi-faceted analysis , or in other words, through the above steps, the extraction and effective representation of the global and local features of the target shape are realized, thus avoiding the low recognition accuracy in the background technology based only on global or local features; and, different Due to the high computational complexity of the target recognition method provided by the background technology, the multi-scale invariant descriptor used in the embodiment of the present invention when matching the target to be recognized with the template usually has a small dimension, so the computational complexity is relatively small. Low, achieving high recognition efficiency and recognition accuracy at the same time.

In addition, while extracting and effectively representing global and local features of the target shape, the present invention has excellent performances such as scale invariance, rotation invariance, hinge invariance, translation invariance, and occlusion invariance, and effectively suppresses noise interference, which further improves the recognition accuracy and recognition efficiency.

In a target recognition method based on multi-scale invariant description provided by an embodiment of the present invention, a closed contour is extracted from the edge of the target shape, which may specifically include:

The Canny operator is used to extract a closed contour for the edge of the target shape.

It should be noted that, when extracting the edge of the target shape, any method predetermined by the staff that can effectively realize the extraction of the edge of the target shape can be used, specifically Canny operator, Laplacian operator, etc., the embodiment of the present invention Among them, the Canny operator is preferred to effectively and quickly realize the extraction of the contour of the target shape.

In an object recognition method based on multi-scale invariant description provided by an embodiment of the present invention, the number of layers of the contour is determined, as shown in FIG. 3 , which may specifically include:

S21: Determine that the current layer is 1.

S22: Add 1 to the current layer as the current layer, calculate the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to the current layer, and the area parameter and arc length of the layer obtained by subtracting 1 from the contour point corresponding to the current layer The difference between the parameter and the center of gravity parameter, and judge whether the ratio of the sum of the differences corresponding to all contour points to the number of contour points is less than the difference threshold, and if so, determine the number of layers corresponding to the layer obtained by subtracting 1 from the current layer as the contour number of layers, if not, go to step S23.

Among them, the difference threshold can be determined according to actual needs. For example, when the current layer is 2, each contour point corresponds to the area parameter, arc length parameter, and center of gravity parameter of the current layer, and when the current layer is 3, the contour point corresponds to the area of the current layer. The average value of the difference between parameters, arc length parameters and center of gravity parameters, that is, the ratio of the sum of the above differences corresponding to all contour points to the number of contour points is less than the difference threshold, then the number of layers of each contour point is determined to be 2 layers . In addition, the average value refers to the average value of the difference corresponding to each contour point, and the difference of any contour point is the difference between the area parameter, arc length parameter and center of gravity parameter corresponding to the contour point in different layers. When determining the difference between the area parameters, arc length parameters and center-of-gravity parameters of any contour point corresponding to the two layers, the difference between the area parameters corresponding to the contour point in the two layers can be calculated respectively. The difference of the arc length parameters corresponding to each layer and the difference of the center of gravity parameters corresponding to the two layers respectively, and then calculate the final difference according to the above three differences and the weights of the three differences, or each The area parameter, arc length parameter, and center of gravity parameter corresponding to the layer constitute a parameter vector, and then calculate the vector difference between the parameter vectors corresponding to the two layers respectively, and obtain the final difference. The above difference can also be calculated according to other methods according to actual needs. All within the protection scope of the present invention.

S23: return to step S22.

In addition, it should be noted that the initial area parameter, arc length parameter and center of gravity parameter of each contour point can also be calculated in advance, that is, the initial area parameter, initial arc length parameter and initial center of gravity of each contour point when the current layer is 1 parameter, because the number of layers of the contour is at least one layer, therefore, the initial area parameter, initial arc length parameter and initial center of gravity parameter of each contour point can be calculated first, and then any contour point can be selected to calculate its The difference between the above parameters corresponding to the current layer and the layer obtained by subtracting one from the current layer, and when the difference is less than the difference threshold, determine the number of layers corresponding to the current layer minus one as the number of layers of the outline, and determine the layer After counting, it is enough to calculate the above-mentioned parameters corresponding to other layers for each contour point. Of course, other settings can also be made according to actual needs, all of which are within the protection scope of the present invention.

In addition, calculate the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to the current layer, which may specifically include:

Determine any contour point as the target contour point, take the coordinates of the target contour point as the center, and make a circle with the radius corresponding to the current layer as the preset radius to obtain the preset circle corresponding to the current layer;

The ratio of the area of the area intercepted by the preset circle in the target shape and the area directly connected with the target contour point to the area of the preset circle is used as the area parameter of the target contour point. The value range of the area parameter should be between 0 and 1 between;

The ratio of the length of the arc segment cut out by the preset circle in the target shape and having a direct connection relationship with the target contour point to the circumference of the preset circle is used as the arc length parameter of the target contour point, and the value range of the arc length parameter Should be between 0 and 1;

Determine the distance between the center of gravity and the target contour point of the area intercepted by the preset circle in the target shape that has a direct connection relationship with the target contour point, and use the ratio of the distance to the preset radius as the center of gravity parameter of the target contour point, and the center of gravity parameter The value range of should be between 0 and 1.

Wherein, it should be noted that the preset radius is the radius corresponding to the current layer, that is, different layers correspond to different preset radii. Moreover, for each contour point, it is necessary to obtain its area parameter, arc length parameter and center-of-gravity parameter corresponding to each layer according to the above steps, which will not be repeated here.

After the preset circle C ₁ (i) is obtained according to the above steps, part of the target shape must fall within the preset circle. Assuming that the target shape is shown in Figure 4, the schematic diagram of the preset circle and the target shape is shown in Figure 5 Show. If the part of the target shape falling within the preset circle is a separate area, then this separate area is the area that has a direct connection relationship with the target contour point, denoted as Z ₁ (i); if the target shape falls within the preset circle If the part of the target contour point is divided into several disconnected regions, such as region A and region B shown in Figure 5, then the region of the target contour point on its contour is determined to be the region that has a direct connection relationship with the target contour point, denoted as Z ₁ (i). Specifically, the area of the area Z ₁ (i) in the preset circle C ₁ (i) that is directly connected to the target contour point p(i) is recorded as Then there are:

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; x</span>}$

Among them, B(Z ₁ (i),x) is an indicator function defined as

Take the ratio of the area of Z ₁ (i) to the area of the preset circle C ₁ (i) as the area parameter s ₁ (i) of the multi-scale invariant descriptor of the target contour point, namely:

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; (</span>{{\mathrm{<span\; class="notranslate">\; \&pi;r</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}$

The value range of s ₁ (i) should be between 0 and 1.

When calculating the center of gravity of an area that is directly connected to the target contour point, the average of the coordinate values of all pixel points in the area can be calculated, and the result is the coordinate value of the center of gravity of the area, which can be expressed as:

${\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{<span\; class="notranslate">\; \&Integral;</span>}_{{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; x</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; x</span>}}{{<span\; class="notranslate">\; \&Integral;</span>}_{{\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; x</span>}}$

Wherein, w ₁ (i) is the center of gravity of the above region.

And calculate the distance between the target contour point and the center of gravity w ₁ (i) It can be expressed as:

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>$

and will The ratio of the radius of the preset circle to the target contour point is used as the barycenter parameter c ₁ (i) of the multi-scale invariant descriptor of the target contour point, namely

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}$

The value range of c ₁ (i) should be between 0 and 1.

After the preset circle is obtained according to the above steps, after the contour of the target shape is cut by the preset circle, there must be one or more arc segments within the preset circle, as shown in FIG. 6 . If the target shape has only one arc segment within the preset circle, then it is determined that the arc segment is an arc segment that has a direct connection relationship with the target contour point. If the target shape has multiple arc segments within the preset circle, as shown in Figure 6 Segment A (SegmentA), arc segment B (SegmentB), arc segment C (SegmentC), then determine that the arc segment where the target contour point is located is an arc segment that has a direct connection relationship with the target contour point, which is the arc segment in Figure 6 A (SegmentA).

The length of the arc segment directly connected with the target contour point p(i) in the preset circle C ₁ (i) is recorded as and will The ratio of the circumference of the preset circle C ₁ (i) is used as the arc length parameter l ₁ (i) of the multi-scale invariant descriptor of the target contour point, namely

${\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&pi;r</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; )</span>}$

Wherein, the value range of l ₁ (i) should be between 0 and 1.

Therefore, the multi-scale invariant descriptors of the target contour point and all other contour points can be obtained by the above method, expressed as M(i):

M(i)＝{s _k (i),l _k (i),c _k (i)|k∈[1,m],i∈[1,n]}

In a target recognition method based on multi-scale invariant description provided by an embodiment of the present invention, determining the radius corresponding to the current layer as the preset radius may specifically include:

The ratio of the equivalent radius of the target shape to 2 to the Nth power is taken as the preset radius corresponding to the current layer, where N is the number of layers corresponding to the current layer.

Specifically, it can be expressed as: taking p(i) as the center and r ₁ as the preset radius to make a circle to obtain the preset circle C ₁ (i), which is used to calculate the multi-scale invariant descriptor of the corresponding contour point Do the preparation work.

The specific representation of the preset radius r ₁ can be:

${\mathrm{<span\; class="notranslate">\; r</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; R</span>}}{{\mathrm{<span\; class="notranslate">\; 2</span>}}^{\mathrm{<span\; class="notranslate">\; N</span>}}}$

Among them, R is the equivalent radius of the target shape, and N is the number of layers corresponding to the current layer. When calculating the preset radius r ₁ corresponding to the first layer, N in this formula is taken as 1; when calculating the preset radius corresponding to other layers, N is the corresponding number of layers.

In addition, determine the equivalent radius of the target shape, which may specifically include:

Calculate the area of the target shape, and take the square root of the area of the target shape to obtain the equivalent radius of the target shape.

Of course, the specific calculation methods of the preset radius and the equivalent radius mentioned above can also be set by the staff according to actual needs, all of which are within the protection scope of the present invention.

Specifically, it can be expressed as:

$\mathrm{<span\; class="notranslate">\; R</span>}<span\; class="notranslate">\; =</span>\sqrt{{\mathrm{<span\; class="notranslate">\; area</span>}}_{\mathrm{<span\; class="notranslate">\; S</span>}}}$

Among them, area _S is the area of the target shape, and R is the equivalent radius of the target shape.

In the object recognition method based on multi-scale invariant description provided by the embodiment of the present invention, the multi-scale invariant descriptor of each contour point is used to match the target to be recognized with the template in the preset template library, and the target to be recognized is obtained. Identify the best matching template for the target, which can include:

Calculate the matching degree between the multi-scale invariant descriptor of the target to be recognized and the multi-scale invariant descriptor of the template, and determine that the template whose matching degree is not greater than that of other templates is the best matching template.

Among them, the smaller the matching degree, the more similar the shape of the object to be recognized is to the shape of the corresponding template. Therefore, it is determined that the template whose matching degree is not greater than that of other templates is the best matching template. And the template whose matching degree is not greater than the matching degree of other templates can be specifically: if there is a template with the smallest matching degree in the template, then determine that the template is the best matching template, if there are multiple templates with the smallest matching degree and equal in the template , then determine one of the templates as the best matching template.

Specifically, the calculation of the matching degree between the multi-scale invariant descriptor of the target to be identified and the multi-scale invariant descriptor of the template may include:

Arranging the contour points of the target to be identified in order to form a target sequence, and arranging the contour points of a template that needs to be matched with the target to be identified in order to form a matching sequence;

A dynamic programming algorithm is used to calculate the matching degree between the target sequence and the matching sequence as the matching degree between the target to be identified and the corresponding template.

Of course, other algorithms can also be pre-set by the staff to calculate the above-mentioned matching degree according to actual needs, all of which are within the protection scope of the present invention.

Wherein, using the dynamic programming algorithm to calculate the matching degree between the target sequence and the matching sequence may include:

Find the Euclidean distance between the multi-scale invariant descriptors corresponding to every two contour points, and determine the distance as the matching cost between the two contour points, where the two contour points are included in the target sequence and the matching sequence;

The sum of the obtained matching costs is taken as the matching degree.

Specifically, the above calculation of the matching degree between the target to be identified and the corresponding template may include:

The contour points belonging to the target sequence and the matching sequence are pairwise matched according to the preset rules, and the matching cost between each pair of contour points is obtained, that is, one of the two contour points paired each time belongs to the target sequence , and the other belongs to the matching sequence. Among them, the preset rules can be determined by the staff according to actual needs, which can be as follows:

1. The two contour points for pairing must belong to two different point sequences, that is, the target sequence and the matching sequence respectively;

2. Contour points that have participated in the matching cannot participate in the matching again;

3. When all the contour points in the point sequence with a small number of contour points have matching contour points, the matching process ends.

Adding the matching cost of each pair of two contour points in this matching is the matching cost corresponding to this matching method. By repeating the matching process of the above contour points multiple times, multiple matching methods can be obtained, and the matching costs of all possible matching methods are calculated, and the minimum matching cost is used as the two point sequences, that is, the target sequence and the matching sequence corresponding to suitability.

Specifically, the target sequence can be expressed as A={p ₁ ,p ₂ ,...,p _m }, and the matching sequence can be expressed as B={q ₁ ,q ₂ ,...,q _n }, without losing In general, it can be assumed that m≥n. Then calculate the Euclidean distance between the multi-scale invariant descriptors of two contour points p _i and q _j belonging to different point sequences, and the obtained value is used as the matching cost d(p _i , q _j ) of these two contour points ,which is:

$\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\sqrt{{<span\; class="notranslate">\; (</span>{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; (</span>{{\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; l</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; (</span>{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}}^{\mathrm{<span\; class="notranslate">\; q</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; \&Element;</span><span\; class="notranslate">\; \&lsqb;</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&rsqb;</span>$

Then the matching cost matrix D generated by pairwise matching of all possible contour points is:

Finally, record the corresponding matching method after matching according to the above principles as π, then the matching cost f _{A, B} (π) corresponding to the matching method π is the sum of the matching costs of each pair of two contour points in this matching and that is:

${\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; \&pi;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; )</span>$

Use the dynamic programming algorithm to calculate the matching method π that makes the matching cost f _A,B (π) the smallest, as the matching degree sim(A,B) between the target sequence and the matching sequence, that is

sim(A,B)=minf _A,B (π)

It should be noted that if the functions of the target recognition method based on multi-scale invariant description provided by the embodiment of the present invention are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computing device readable storage medium. Based on this understanding, the part of the embodiment of the present invention that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to make a A computing device (which may be a personal computer, a server, a mobile computing device or a network device, etc.) executes all or part of the steps of the methods in various embodiments of the present invention. The aforementioned storage medium may include various media capable of storing program codes such as U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk.

Specifically, corresponding to the above-mentioned method embodiment, the embodiment of the present invention also provides a target recognition device based on multi-scale invariant description, as shown in FIG. 7 , which may include:

The extraction module 1 is used to obtain the shape of the target to be recognized as the target shape, and extract a closed contour from the edge of the target shape, and obtain all contour points on the contour and the coordinates of each contour point;

Calculation module 2, used to determine the number of layers of the contour, and calculate the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to each layer based on the coordinates of each contour point, as the multi-scale invariant description of the contour point son;

The matching module 3 is used to use the multi-scale invariant descriptor of each contour point to match the target to be recognized with the template in the preset template library, obtain the best matching template of the target to be recognized, and determine the best matching template The category of is the category of the target to be recognized.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the extraction module may include:

The extracting unit is used for extracting a closed contour aiming at the edge of the target shape by using the Canny operator.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the calculation module may include:

Determine the unit to perform the following steps:

Step A: Determine that the current layer is 1;

Step B: Add 1 to the current layer as the current layer, calculate the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to the current layer, and the area parameter and arc of the layer obtained by subtracting 1 from the contour point corresponding to the current layer The difference between the length parameter and the center of gravity parameter, and determine whether the ratio of the sum of the differences corresponding to all contour points to the number of contour points is less than the difference threshold, and if so, determine the number of layers corresponding to the layer obtained by subtracting 1 from the current layer as the contour The number of layers, if not, go to step C;

Step C: Go back to step B.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the calculation module may include:

The calculation unit is used to: determine any contour point as the target contour point, take the coordinates of the target contour point as the center, and take the radius corresponding to the current layer as the preset radius to make a circle to obtain the preset circle corresponding to the current layer; The ratio of the area of the area that is intercepted by the preset circle in the target shape and has a direct connection relationship with the target contour point to the area of the preset circle is used as the area parameter of the target contour point; the target shape is cut out by the preset circle, The ratio of the length of the arc segment that has a direct connection relationship with the target contour point to the circumference of the preset circle is used as the arc length parameter of the target contour point; if the target shape is intercepted by the preset circle, it has a direct connection relationship with the target contour point The distance between the center of gravity of the region and the target contour point, and the ratio of the distance to the preset radius is used as the center of gravity parameter of the target contour point.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the calculation unit may include:

The radius determination unit is configured to use the ratio of the equivalent radius of the target shape to 2 to the Nth power as the preset radius corresponding to the current layer, where N is the number of layers corresponding to the current layer.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the radius determining unit may include:

The radius determining subunit is used for calculating the area of the target shape, and taking the square root of the area of the target shape to obtain the equivalent radius of the target shape.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the matching module may include:

The matching unit is used to calculate the matching degree between the multi-scale invariant descriptor of the target to be identified and the multi-scale invariant descriptor of the template, and determine that the template whose matching degree is not greater than that of other templates is the best matching template.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the matching unit may include:

The matching subunit is used to: arrange the contour points of the target to be identified in order to form a target sequence, arrange the contour points of a template that needs to be matched with the target to be identified in order to form a matching sequence; use a dynamic programming algorithm to calculate the target sequence and The matching degree between matching sequences is used as the matching degree between the target to be identified and the corresponding template.

In an object recognition device based on multi-scale invariant description provided by an embodiment of the present invention, the matching subunit may include:

The obtaining unit is used to: obtain the Euclidean distance between the multi-scale invariant descriptors corresponding to every two contour points, and determine the distance as the matching cost between the two contour points, wherein the two contour points Included in the target sequence and the matching sequence respectively; the sum of the obtained matching costs is taken as the matching degree.

For the description of the relevant parts of the target recognition device based on the multi-scale invariant description provided by the embodiment of the present invention, please refer to the detailed description of the corresponding part in the target recognition method based on the multi-scale invariant description provided by the embodiment of the present invention. I won't repeat them here.

In addition, each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same or similar parts of each embodiment can be referred to each other.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. A target recognition method described based on multi-scale invariants, characterized in that, comprising: Obtaining the shape of the target to be identified as the target shape, and extracting a closed contour from the edge of the target shape, obtaining all contour points on the contour and the coordinates of each contour point; Determining the number of layers of the contour, and calculating the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to each layer based on the coordinates of each contour point, as the multi-scale invariant descriptor of the contour point; Using the multi-scale invariant descriptor of each contour point, match the target to be recognized with the template in the preset template library to obtain the best matching template of the target to be recognized, and determine the best The category of the matching template is the category of the target to be identified. 2. method according to claim 1, is characterized in that, extracts a closed contour by the edge of described target shape, comprises: A Canny operator is used to extract a closed contour for the edge of the target shape. 3. The method according to claim 1, wherein determining the number of layers of the outline comprises: Step A: Determine that the current layer is 1; Step B: add 1 to the current layer as the current layer, calculate the area parameter, arc length parameter and center of gravity parameter corresponding to each of the contour points corresponding to the current layer and the area parameter of the layer obtained by subtracting 1 from the contour point corresponding to the current layer , the difference between the arc length parameter and the center of gravity parameter, and judge whether the ratio of the sum of the differences corresponding to all the contour points to the number of the contour points is less than the difference threshold, if so, then determine the current layer minus 1 to obtain The layer number corresponding to the layer is the layer number of the outline, if not, then perform step C; Step C: Go back to step B. 4. The method according to claim 3, wherein calculating each of the contour points corresponding to the area parameter, arc length parameter and center of gravity parameter of the current layer comprises: Determining any contour point as a target contour point, taking the coordinates of the target contour point as the center, and making a circle with a radius corresponding to the current layer as a preset radius, to obtain a preset circle corresponding to the current layer; Taking the ratio of the area of the area intercepted by the preset circle in the target shape and having a direct connection relationship with the target contour point to the area of the preset circle as the area parameter of the target contour point; Taking the ratio of the length of the arc segment cut out by the preset circle in the target shape and having a direct connection relationship with the target contour point to the circumference of the preset circle as the arc of the target contour point long parameter; Determining the distance between the center of gravity of the region intercepted by the preset circle in the target shape and having a direct connection relationship with the target contour point and the target contour point, and calculating the ratio of the distance to the preset radius As the center of gravity parameter of the target contour point. 5. The method according to claim 4, wherein determining that the radius corresponding to the current layer is a preset radius comprises: The ratio of the equivalent radius of the target shape to 2 to the Nth power is used as the preset radius corresponding to the current layer, where N is the number of layers corresponding to the current layer. 6. The method according to claim 5, wherein determining the equivalent radius of the target shape comprises: calculating the area of the target shape, and taking the square root of the area of the target shape to obtain an equivalent radius of the target shape. 7. The method according to claim 1, characterized in that, using the multi-scale invariant descriptor of each of the contour points, the target to be identified is matched with a template in a preset template library to obtain the The best matching template for the target to be identified, including: Calculating the matching degree between the multi-scale invariant descriptor of the target to be identified and the multi-scale invariant descriptor of the template, and determining the template whose matching degree is not greater than that of other templates as the best matching template . 8. The method according to claim 7, wherein the calculation of the matching degree between the multi-scale invariant descriptor of the target to be identified and the multi-scale invariant descriptor of the template comprises: Arranging the contour points of the target to be identified in order to form a target sequence, and arranging the contour points of a template that needs to be matched with the target to be identified in order to form a matching sequence; Using a dynamic programming algorithm to calculate the matching degree between the target sequence and the matching sequence as the matching degree between the target to be identified and the corresponding template. 9. The method according to claim 8, wherein the matching degree between the target sequence and the matching sequence is calculated using a dynamic programming algorithm, comprising: Calculate the Euclidean distance between the multi-scale invariant descriptors corresponding to every two contour points, and determine the distance as the matching cost between the two contour points, wherein the two contour points are respectively included in said target sequence and said matching sequence; The sum of the obtained matching costs is used as the matching degree. 10. A target recognition device based on multi-scale invariant description, characterized in that it comprises: The extraction module is used to obtain the shape of the target to be identified as the target shape, and extract a closed contour from the edge of the target shape, and obtain all contour points on the contour and the coordinates of each contour point; Calculation module, used to determine the number of layers of the contour, and calculate the area parameter, arc length parameter and center of gravity parameter of each contour point corresponding to each layer based on the coordinates of each contour point, as the multi-scale of the contour point invariant descriptor; A matching module, configured to use the multi-scale invariant descriptor of each contour point to match the target to be identified with templates in a preset template library to obtain the best matching template of the target to be identified, and Determining the category of the best matching template as the category of the target to be identified.